Wang Duomin

icon indicating a website link https://dorniwang.github.io/ icon indicating an email [email protected]

icon company stepfun icon location Shanghai icon location I'm currently a researcher at Stepfun. I'm focusing on body and portrait driving, multi-modal model and representation lerning.

Results 6 comments of Wang Duomin

formulation error

yes，i found this and confuse about it too

about rasterization for T_uv

I found that you obtain texture-rasterized tri-planes of each orthogonal views by summation two opposite view's result. But how to blend T_uv and T_static since the alpha masks are different...

为什么3D VAE的结果仔细看是由64*64的网格组成的？

> Is this what you are finding? > > https://github.com/hpcaitech/Open-Sora/blob/476b6dc79720e5d9ddfb3cd589680b2308871926/opensora/models/layers/blocks.py#L79 This PatchEmbed3D is used to process the output of 3DVAE Encoder's output and then feed the patched latents into Dit...

Pose 图 F_t 到底是怎么渲染的？How do you render the pose image?

我理解这是一个rasterization的过程。至于骨骼点可以绑定smpl然后rasterize成这个特征图

Pose 图 F_t 到底是怎么渲染的？How do you render the pose image?

> > 我理解这是一个rasterization的过程。至于骨骼点可以绑定smpl然后rasterize成这个特征图 > > which one, the former or the latter? The rasterization result is the former one, which is shown in the pipeline figure.

Pose 图 F_t 到底是怎么渲染的？How do you render the pose image?

> > 我理解这是一个rasterization的过程。至于骨骼点可以绑定smpl然后rasterize成这个特征图 > > yes, the pose representation is a interpolated feature map via the rasterization, visualized as the former one. I have another question, which pretrained repose model...